Mirror Neurons Responding to Observation of Actions Made with Tools in Monkey Ventral Premotor Cortex

Pier Francesco Ferrari, Stefano Rozzi, and Leonardo Fogassi Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/17/2/212/1757134/0898929053124910.pdf by guest on 18 May 2021

Abstract & In the present study, we describe a new type of congruent in that they share the same general goal, that is, visuomotor neurons, named tool-responding mirror neurons, taking possession of an object and modifying its state. It is which are found in the lateral sector of monkey ventral hypothesized that after a relatively long visual exposure to premotor area F5. Tool-responding mirror neurons discharge tool actions, a visual association between the hand and the when the monkey observes actions performed by an tool is created, so that the tool becomes as a kind of experimenter with a tool (a stick or a pair of pliers). This prolongation of the hand. We propose that tool-responding response is stronger than that obtained when the monkey mirror neurons enable the observing monkey to extend observes a similar action made with a biological effector (the action-understanding capacity to actions that do not strictly hand or the mouth). These neurons respond also when the correspond to its motor representations. Our findings support monkey executes actions with both the hand and the mouth. the notion that the motor cortex plays a crucial role in The visual and the motor responses of each neuron are understanding action goals. &

INTRODUCTION internal motor representation of the same action in the In area F5 of monkey ventral premotor cortex, there are observer. neurons that become active both when the monkey Although many mirror neurons are activated by the executes hand and mouth goal-directed actions and observation of actions made with a specific effector, a when it observes similar actions made by another in- considerable number of F5 mirror neurons are activated dividual (Ferrari, Gallese, Rizzolatti, & Fogassi, 2003; by the observation of actions independent of the effec- Gallese, Fadiga, Fogassi, & Rizzolatti, 1996; Rizzolatti, tor (see Ferrari, Fogassi, Gallese, & Rizzolatti, 2001). For Fadiga, Gallese, & Fogassi, 1996). These neurons have example, a mirror neuron may respond when the mon- been called mirror neurons (Gallese et al., 1996). Mirror key observes another individual breaking a peanut with neurons do not respond either to simple object presen- the hand and also when the same action is performed tation or to vision of an action mimed with the hand or with the mouth. Thus, mirror neurons have the property with the mouth. Observation of actions performed using to generalize the meaning of an observed action inde- a tool, such as pliers, is not effective in eliciting mirror pendently of its specific visual features. neuron response. The visual response of mirror neurons Brain imaging and magnetic encephalogram studies is largely independent from the distance at which the have demonstrated in humans a system for action observed action is performed and from the properties of recognition similar to that of monkeys. It includes the the object’s target of the observed actions. Most impor- premotor cortex and the adjacent area 44, Broca’s area tantly, mirror neurons show a very good congruence (Buccino, Binkofski, et al., 2001; Nishitani & Hari, 2000; between the effective observed and the effective execut- Iacoboni et al., 1999; Grafton, Arbib, Fadiga, & Rizzolatti, ed action. This visuomotor congruence has prompted 1996; Rizzolatti, Fadiga, Matelli, et al., 1996), the human the idea that the basic function of mirror neurons con- homologue of area F5 (Petrides & Pandya, 1994; von sists in understanding actions made by other individuals Bonin & Bailey, 1947). However, differently from (Rizzolatti, Fogassi, & Gallese, 2000; Gallese et al., 1996; monkeys, fMRI studies showed that the human inferior Rizzolatti, Fadiga, Gallese, et al., 1996; di Pellegrino, frontal cortex is activated also by observation of mimed Fadiga, Fogassi, Gallese, & Rizzolatti, 1992) by a process actions (Buccino, Binkofski, et al., 2001; Gre`zes, Costes, that matches the visual description of an action with the & Decety, 1998). This latter finding is in agreement with the common observation that humans interpret move- ments as goal-directed also when there is no target (mimed actions) or when the target is present, but the Universita` di Parma, Italy action is made with a tool.

D 2005 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 17:2, pp. 212–226 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/0898929053124910 by guest on 30 September 2021 The experiments reported above are evidence that actions. The large prevalence of mouth-movement- action understanding in humans is linked to an activa- related neurons is most likely because of the lateral lo- tion of the mirror system, also when understanding cation of most penetrations. implies a high level of generalization. Apparently, this Of all recorded neurons, 143 (68.4%) responded to latter property seems absent in monkeys as far as mimed visual stimuli. This percentage is relatively high when actions and actions made with tools are concerned. compared with previous work from our laboratory on However, few observations made in our laboratory area F5. This is probably because of the fact that during showed that at the end of a relatively long period of recording the focus of our interest was concentrated on experiments, it was possible to find mirror neurons neurons with visual responses and in particular on those responding also to actions made by the experimenter responding to biological stimulation. Forty-two neurons Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/17/2/212/1757134/0898929053124910.pdf by guest on 18 May 2021 with tools (see Arbib & Rizzolatti, 1999). (20.1% of all recorded neurons) responded best to the In the present study, we describe the properties of a observation of actions performed with a tool by an new class of F5 visuomotor neurons (tool-responding experimenter. Because of the similarity of their visuo- mirror neurons) responding to the observation of ac- motor properties with those of mirror neurons, these tions performed with a tool. In accord with the above- neurons will be referred to as tool-responding mirror mentioned observations, all neurons reported in the neurons. They do not respond to simple presentation of present study were found after a long period of experi- the tool or of other objects or food. All tool-responding ments, during which the monkeys had been highly mirror neurons had a motor response. Of 42 tool- exposed to actions made with tools by the experiment- responding mirror neurons, 8 (19.0%) responded during ers. Considering the importance of the idea of general- the execution of actions made with the hand, 3 (7.2%) ization of action goal at the single neuron level, the aim with the mouth, and 31 (73.8%) with both the hand and of this study was that of verifying in a quantitative way the mouth. the properties of this new class of neurons and their Of the other visual neurons, 74 (35.4% of the total differences with hand mirror neurons. number of recorded neurons) were typical mirror neu- rons responding best to the observation of hand, mouth, or hand and mouth actions. A subset of these mirror RESULTS neurons (n = 12) presented also a weaker response to observation of actions performed with a tool. The Properties of Recorded Neurons remaining visual neurons (n = 27, 12.9% of the total We recorded a total of 209 neurons from area F5 of two number of recorded neurons) responded to other types macaque monkeys (Macaca nemestrina). One hun- of visual stimuli, such as object presentation or object dred forty-one neurons were recorded from Monkey motion. 1 (right hemisphere) and 68 from Monkey 2 (left hemisphere). Most recorded neurons were located in Tool-Responding Mirror Neurons Properties the lateral part of area F5, as defined by Matelli, Luppino, and Rizzolatti (1985). In Monkey 1, we inves- Thirty-three tool-responding mirror neurons were stud- tigated also part of medial F5. ied long enough to quantitatively assess their proper- All neurons were tested for their responsiveness to ties. The specificity of the neuron discharge was monkey active movements and to visual stimuli (see also assessed by comparing its response during the condi- Methods). Testing motor properties mainly consisted in tion in which the monkey observed an action made eliciting hand, arm, and mouth actions by presenting with a tool, with the response during the other con- to the monkey pieces of food and objects of different ditions as, namely, observation of actions made with a sizes, shapes, and orientations, introduced in every biological effector, observation of mimed actions per- spatial quadrant. Testing of visual properties consisted formed with a tool, and simple object observation. All in the presentation of actions made by the experimenter tool-responding mirror neurons had higher responses with the hand, the mouth, and a tool. The tools used in to observation of actions made with a stick (n = 26), this study were chosen based on the similarity between with pliers (n =4),orwithboth(n =3)when the types of goal that could be achieved with them and compared with actions made with a biological effector, the types of goal normally achieved with the biological either the hand or the mouth ( p <.05,two-way effector. Two types of tools were employed: a stick with analysis of variance [ANOVA] for repeated measures a metal tip and a pair of pliers. Other visual stimuli con- followed by a post hoc test). These neurons did not sisted of the presentation of static and moving objects respond to observation of actions mimed with the tool at different space locations. or with a biological effector without the target object. All recorded neurons, but 8 (3.8%), had motor prop- Similarly, they did not respond to simple presentation erties. More specifically, 52 (24.9%) discharged in asso- of an object when this was presented on the tip of the ciation with active hand actions, 40 (19.1%) with mouth tool. When tested, tool-responding mirror neurons did actions, and 109 (52.2%) with both hand and mouth not respond to the observation of actions made with an

Ferrari, Rozzi, and Fogassi 213 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/0898929053124910 by guest on 30 September 2021 unfamiliar tool (a signal stick, which is a wooden, without food elicited only a very weak response (C). colored circle attached to the tip of a long stick) aimed No response was elicited by the simple presentation of to touch the target object, but devoid of the possibility a piece of food held on the stick (D). A strong to pick it up. The visual response of tool-responding response was evoked when the monkey grasped a mirror neurons was always excitatory. Note that in small piece of food with its hand (E) or its mouth 29 neurons, there was also a weak response to obser- (F). Note that also the motor response, as the visual vation of an action made with the biological effector, response, began during the approaching phase of whereas in 4 neurons, this response was absent. grasping and peaked when the hand or the mouth The responses of tool-responding mirror neurons closed on the food. were independent of the distance and the space sector Figure 2 shows an example of a neuron (Unit 102) Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/17/2/212/1757134/0898929053124910.pdf by guest on 18 May 2021 where the observed action was performed by the selective for the observation of an action made with experimenter. Similarly, the size and shape of the pliers. The neuron discharged when the experimenter object target did not seem to affect the intensity of grasped and broke a peanut, held on his hand, using a the response, provided that the differences of these pair of pliers (A). The response occurred during both physical properties were not such as to modify the grasping and breaking phases. The same grasping and meaning of the action. For example, sticking a small or breaking actions performed by the experimenter with a large piece of apple evoked the same discharge. his hand were much less effective (B). This neuron Although not tested in all neurons, the direction of responded also when the monkey broke a small piece the action toward the target did not seem to influence of food with the hand (C) or the mouth (D). Note that the neuron response. the intensity of neuron discharge during monkey break- Table 1 summarizes the main categories of ‘‘tool- ing action is lower than that during action observation, responding mirror neurons’’ (see Methods section for probably because the breaking action made by the the definition of the types of actions made with the stick monkey is more prolonged in time. and the pliers). The classification criterion was based on Figure 3 shows an example of a neuron (Unit 111) the most effective observed action performed with a tool. selective for the observation of an action made with The most represented categories, either alone or associ- pliers. The neuron discharged when the experimenter ated to other actions, were ‘‘sticking’’ and ‘‘holding.’’ grasped and held a piece of food, placed on a tray, using Figure 1 shows an example of a neuron (Unit 88) a pair of pliers moved with his right hand (A). The selective for the observation of a sticking action. The response occurred during both grasping and holding experimenter approached a small piece of food placed phases. The same grasping and holding actions per- on one hand with a stick held with the other, took the formed by the experimenter with the stick (B) was less food, and held it (Condition A). This neuron began to effective, in particular during the approaching phase respond during the approaching phase of the stick to until the contact of the tool with the food. During the food. The response peak occurred when the food observation of grasping action made with the hand (C), was punctured, and the discharge continued during the neuron response was almost absent. This neuron the holding phase. The neuron response was smaller responded also when the monkey grasped and held a when the experimenter grasped the same piece of piece of food with the hand (D). food with his hand (B). Mimicking the sticking action Figure 4 shows an example of a neuron (Unit 95) responding to the observation of actions made on a piece of food with both stick (A) and pliers (B). A Table 1. Tool-Responding Mirror Neurons Subdivided strong discharge was present during the approaching, According to the Most Effective Observed Actions in grasping, and holding phase with both tools. Statistical Activating Them analysis revealed that the response to observation of Category Number of Neurons the action performed with the stick was greater than that performed with the pliers. In contrast, the obser- Sticking 5 vation of a grasping action performed with the hand Holding 4 was not effective (C). Miming a sticking action was also poorly effective (D). Finally, the strongest response Others 4 was found when the monkey grasped food with the Sticking or grasping*/holding 13 hand (E). Sticking/manipulating 2 In general, most ‘‘tool-responding mirror neurons’’ differentiated between observed actions made with tools Approaching/sticking 2 and those made with biological effectors, as expressed Others 3 by the quantitative significant difference in the discharge displayed in the two different conditions. In few cases Total 33 (n = 2), this difference was further strengthened by a *The term grasping is referred to action made with pliers. completely opposite excitatory/inhibitory behavior of

214 Journal of Cognitive Neuroscience Volume 17, Number 2 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/0898929053124910 by guest on 30 September 2021 the neuron discharge. A clear example of this type of neurons is presented in Figure 5. This neuron (Unit 92) was selective for the observation of sticking a piece of food. The excitatory response (A) was observed from the moment in which the experimenter approached the food with the stick to the whole holding phase. In contrast, when the experimenter grasped food with the hand (B), during the approaching and grasping phase, there was a complete inhibition of the neuron response. However, the holding phase, similarly to Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/17/2/212/1757134/0898929053124910.pdf by guest on 18 May 2021 condition A, was excitatory. Thus, the discrimination between the two actions is present during the approach- ing and grasping/sticking phase only. When the experi- menter mimed the sticking action (C) and when the food was simply presented on the stick (D), the neuron did not fire. An excitatory motor response occurred when the monkey grasped a small piece of food (not shown in the figure). The capacity of these neurons to discriminate be- tween actions made with the tool and actions made with a biological effector is evident also when the two observation conditions are presented in alternation dur- ing the same acquisition. The neuron (Unit 100) shown in Figure 6 is selective for the observation of an action made with a stick, the response occurring during the holding phase (A). When the experimenter grasped food with the hand (B), there was only a weak neuron response. The same difference in response between the two conditions is clear also when the two visual stimuli are presented in alternation (C). Considering the clear difference, at least in pictorial terms, between the effectors involved in the observed and the executed actions (i.e., tool and hand, respec- tively), we analyzed whether it was possible to find a relationship between the visual and motor response of each neuron in terms of action goal. For example, if a neuron responded to the observation of sticking and holding a piece of food and when the monkey grasped

Figure 1. Example of a tool-responding mirror neuron responding to a sticking action. In each panel, the rasters and the histogram represent the neuron response during a single experimental condition. The histogram represents the average of 10 trials. Rasters and histograms are aligned with the moment in which the stick held by the experimenter touched the food or the tray (observation conditions), when the monkey touched the food with the hand or the mouth (motor conditions), or when the food was abruptly presented (presentation condition). Ordinates, spikes per sec; abscissa, time; bin width, 20 msec. (A) The experimenter approaches with the stick, held in his hand, a piece of food placed on a tray, and then punctures and holds it. (B) The experimenter grasps with his hand a piece of food placed on a tray and then holds it. (C) The experimenter makes the same approaching and sticking movement as in (A), but no food is present. (D) The experimenter introduces abruptly a piece of food held on a stick in the monkey visual field and then keeps it still. In all observation conditions, the stimulus was presented at about 75 cm from the monkey. (E) Monkey grasps a piece of food with the hand. (F) The experimenter moves a piece of food toward the monkey’s mouth; the monkey grasps it with its teeth and eats it.

Ferrari, Rozzi, and Fogassi 215 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/0898929053124910 by guest on 30 September 2021 rons (30/33) show a very good similarity between the goal of the observed and executed effective actions. In several neurons, the similarity was very strict. Thus, the visual and motor responses of tool-responding mirror neurons can be interpreted as sharing the same type of action code. Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/17/2/212/1757134/0898929053124910.pdf by guest on 18 May 2021

Figure 2. Example of a tool-responding mirror neuron responding to an action made with pliers. (A) The experimenter approaches, with a pair of pliers held in his hand, a peanut held on the other hand, then he closes the pliers on the peanut and breaks it. (B) The experimenter grasps with his hand a peanut held on the other hand and then breaks it. (C) Monkey breaks a piece of food with the hand. (D) The experimenter moves a piece of food to the monkey’s mouth; the monkey grasps it with its teeth, breaks it, and eats it. Other conventions as in Figure 1.

and held a similar piece of food, the two responses were Figure 3. Another example of a tool-responding mirror neuron considered similar in terms of action goal, as both responding to an action made with pliers. (A) The experimenter responses signaled taking and keeping possession of approaches, with a pair of pliers held in his hand, a piece of food food. Differently, if a neuron responded to the observa- placed on a tray, then he closes the pliers on the piece of food, grasps tion of taking away food from a tray with the stick and it, and holds it. (B) The experimenter approaches with the stick, held in his hand, a piece of food placed on a tray, and then punctures and during grasping execution by the monkey, the two holds it. (C) The experimenter grasps with his hand a piece of food responses were considered not similar. From this anal- placed on a tray and then holds it. (D) Monkey grasps a piece of food ysis, it emerges that most tool-responding mirror neu- with the hand and holds it. Other conventions as in Figure 1.

216 Journal of Cognitive Neuroscience Volume 17, Number 2 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/0898929053124910 by guest on 30 September 2021 Behavioral Observation wooden platform (1 Â 1 m) was attached to the monkey’s home cage at floor level. Small pieces of food At the end of the recording sessions, we tested in one of (raisins, apple, and peanuts) were placed on the plat- the recorded monkeys (Monkey 1) the ability to use the form at a distance that did not allow the monkey to familiar tool (the stick used as a visual stimulus during reach the food with its hand. Ten minutes after food the experiment) to stick food out of reach. A small presentation, in which the monkey made some unsuc- cessful attempts to take the food, we introduced the stick, placing it on the platform at a reaching distance. The experimenter went out of the monkey room and monitored for 1 hr the monkey’s behavior watching a Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/17/2/212/1757134/0898929053124910.pdf by guest on 18 May 2021 screen connected to a video camera placed 2 m away in front of the monkey. The videotape was then analyzed by two experimenters. During this brief test, the monkey never attempted to use the tool for reaching food, although in the first minutes after the stick was available, the monkey grasped it and bit it. After few minutes, the monkey ignored the stick, trying to take the food by moving the platform.

Histology and Anatomical Location of Tool-Responding Mirror Neurons Figure 7 shows the location of the penetrations where tool-responding mirror neurons were found. Histologi- cal analysis of the microelectrode tracks showed that most penetrations were located in the rostrolateral sector of the post-arcuate cortex. This sector is included within the limits of area F5, as defined based on cyto- chrome oxidase method (Matelli et al., 1985). It should be noted, however, that the cytochrome oxidase histo- chemical material was not available in this case. In fact, because the experimental monkeys were also used for neuroanatomical experiments in which fluorescent dyes were injected in other cortical areas, a fixation protocol was used (paraformaldehyde 3.5%), which, in our expe- rience, gives very poor results in cytochrome oxidase histochemistry. Thus, our penetrations have been as- signed to lateral F5 based on their anatomical location. Note, however, that in Monkey 1, about one third of penetrations were located in the medial part of F5 (see Figure 7A). In this monkey, we made a comparison between the percent of tool-responding mirror neurons recorded in lateral F5 and those recorded in medial F5. In medial F5, 11.8% (4/34) of the recorded neurons were tool-responding mirror neurons, while in lateral F5, the

Figure 4. Example of a tool-responding mirror neuron responding to an action made with both the stick and the pliers. (A) The experimenter approaches with the stick, held in his hand, a piece of food placed on a tray, and then punctures and holds it. (B) The experimenter approaches with the pliers, held in his hand, a piece of food placed on a tray, and then grasps it and holds it. (C) The experimenter grasps with his hand a piece food placed on a tray and then holds it. (D) The experimenter makes the same approaching and sticking movement as in (A), but no food is present. (E) Monkey grasps a piece of food with the hand. Other conventions as in Figure 1.

Ferrari, Rozzi, and Fogassi 217 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/0898929053124910 by guest on 30 September 2021 Time of Exposure to the Acting Tools, Period of Recording, and Frequency of Tool-Responding Mirror Neurons During the training period preceding recording experi- ments, both monkeys were exposed, among other visual stimuli, to the tools used in the present experiment (see also Methods). Recordings lasted 9 months in Monkey 1 and 5.5 months in Monkey 2, with a frequency of 3–4

days of recording per week. The data of the recording Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/17/2/212/1757134/0898929053124910.pdf by guest on 18 May 2021 period before discovery of tool-responding mirror neu- rons have been used for other purposes. Note, however, that also in this period, both monkeys were exposed to tools. The period during which tool-responding mirror neurons have been collected (that is the last period of recording) was the same in both monkeys (1.5 months). In this period, because recording was focused on the acquisition of tool-responding mirror neurons, both monkeys were certainly more exposed to tools than in the previous recording period, although this cannot be quantified. The number of penetrations performed in the lateral part of F5 in the period in which tool- responding mirror neurons have been found was the same in the two monkeys (n = 18). The number of penetrations where tool-responding mirror neurons were found in the lateral part of F5 was significantly higher in Monkey 1 than in Monkey 2 (15/18 in Monkey 1, 6/18 in Monkey 2; p < .01, Fisher’s Exact Probability test). The frequency of tool-responding mirror neurons in Monkey 1 was significantly higher than in Monkey 2 (Z = À3.10, p < .005, Mann–Whitney test). In Monkey 1, a tendency to increase the frequency of occurrence of tool-responding mirror neurons (ex- pressed as total number of tool-responding mirror neu- rons per penetration) was observed in the last part of the recording sessions. In fact, subdividing the total number of penetrations into three equivalent, tempo- rally subsequent, blocks, there was a significant differ- ence between the frequency of tool-responding mirror neurons (out of all recorded neurons) recorded in the third, last block and that of those recorded in the second block (15/36 vs. 7/34, p < .05, Fisher’s Exact Probability test), whereas the frequency of the second Figure 5. Example of a tool-responding mirror neuron responding to a sticking action. (A) The experimenter approaches with the stick, block was not significantly different from that of the first held in his hand, a piece of food placed on a tray, and then punctures block. In Monkey 2, we did not observe any change in and holds it. (B) The experimenter grasps with his hand a piece of the frequency of tool-responding mirror neurons over food placed on a tray and then holds it. (C) The experimenter makes time. This could be partly because of the lower number the same approaching and sticking movement as in (A), but no food of tool-responding mirror neurons found in this monkey is present. (D) The experimenter introduces abruptly a piece of food held on a stick in the monkey visual field and then keeps it still. in respect to the other. Other conventions as in Figure 1.

DISCUSSION percent of tool-responding mirror neurons was 31.3% The present study describes for the first time neurons of (26/83). As far as the depth location of tool-responding a premotor area (F5) responding to the observation of mirror neurons is concerned, almost all of them (more actions made with tools. These visual responses are not than 90%) were recorded at a depth ranging from 200 to evoked by the simple motion of the tool, but by the 2200 Am from the cortical surface (Figure 7B). observation of the interaction between the tool and an

218 Journal of Cognitive Neuroscience Volume 17, Number 2 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/0898929053124910 by guest on 30 September 2021 object. In this respect, these responses resemble the authors (Preuss & Goldman-Rakic, 1991; Barbas & visual responses of typical F5 hand mirror neurons, Pandya, 1987; Vogt & Vogt, 1919), the lateral part of which have been found in the medial part of area F5. the post-arcuate cortex is considered as a subsector The presence in F5 of neurons responding to the partially different from the medial-most sector, thus, observation of tool actions, although not previously suggesting that the histochemically defined area F5 quantitatively described, is not completely unexpected, could be inhomogeneous. It must also be noted that because, as mentioned in the Introduction, in some for Roberts and Akert (1963), the region lying behind mirror neurons, visual responses to actions made with the most lateral part of the inferior arcuate sulcus tools were found after a long period of experiment and would belong to the opercular cortex (PrCO). Although of repeated visual exposure to the observation of the we are more inclined to think that tool-responding Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/17/2/212/1757134/0898929053124910.pdf by guest on 18 May 2021 experimenter using tools (see Arbib & Rizzolatti, 1999). mirror neurons were located inside the border of Consistent with this observation, also in our study, tool- ventral premotor cortex, we cannot completely exclude responding mirror neurons were found during the later that some of the most lateral penetrations were located stage of the experiment. in another anatomical subdivision. A comparison between the present and the previous As far as the homogeneity of area F5 is concerned, studies shows a difference in the location of different functional data suggest that although the medial and the types of mirror neurons. In the present study, tool- lateral sectors of F5 have similarities, they also have responding mirror neurons have been found mainly in partially different visual and motor properties. The the lateral sector of the post-arcuate cortex, whereas similarities between the medial and the lateral sectors the older studies on hand mirror neurons were mainly consist, on the motor side, in the presence of neurons performed in its medial sector (Gallese et al., 1996; coding goal-directed actions and, on the visuomotor Rizzolatti, Fadiga, Gallese, et al., 1996; di Pellegrino side, in the presence of neurons responding to the et al., 1992). According to the histochemical parcella- observation and execution of similar actions (mirror tion of Matelli et al. (1985), the lateral sector of the neurons). However, in the medial sector mirror neurons post-arcuate cortex up to, and often beyond, the in- respond mostly to the observation of hand actions ferior tip of the arcuate sulcus is considered as part of (Gallese et al., 1996), whereas in the lateral sector, most area F5. Based on this parcellation, all our penetrations mirror neurons respond to observation of mouth actions appear to belong to area F5. Note that area F5 has (Ferrari et al., 2003). In the medial-most sector, the hand always been included in the cytoarchitectural area 6 motor representation prevails, whereas in the lateral of Brodmann. The cytoarchitectonic studies on ventral sector there is a predominance of mouth-related motor premotor cortex define the lateral part of the post- neurons (Rizzolatti et al., 1988). In addition to these arcuate cortex until the inferior tip of the arcuate sul- functional distinctions, it is worth noting that in the cus as part of the agranular cortex (Brodmann, 1905) lateral sector of F5 there are also several motor neurons or as a dysgranular cortex (Preuss & Goldman-Rakic, that activate during actions made with both mouth and 1991; Barbas & Pandya, 1987; von Bonin & Bailey, hand (see Ferrari et al., 2001, and the present study), 1947; Vogt & Vogt, 1919). According to some of these suggesting that neurons of this sector could be endowed

Figure 6. Example of a tool-responding mirror neuron responding to a holding action made with a stick. (A) The experimenter approaches with the stick, held in his hand, a piece of food placed on a tray, and then punctures and holds it. (B) The experimenter grasps with his hand a piece food placed on a tray and then holds it. (C) The two conditions (A and B) are presented in alternation with the tool (left) and with the hand (right). Numbers at the side of each panel indicate the number of each trial. Other conventions as in Figure 1.

Ferrari, Rozzi, and Fogassi 219 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/0898929053124910 by guest on 30 September 2021 Figure 7. (A) Lateral view of the 3-D reconstruction of the right hemisphere of Monkey 1. The rectangle reported on the brain reconstruction on the left panel, enlarged on the right panel, includes the part of the ventral premotor cortex explored in the present study. Black circles indicate the penetrations in which Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/17/2/212/1757134/0898929053124910.pdf by guest on 18 May 2021 tool-responding mirror neurons were recorded, white circles the penetrations in which no tool-responding mirror neurons were found. Vertical lines represent the level of three coronal sections, whose outlines are reported in (B). For the classification of the areas of the agranular frontal cortex, see the work of Matelli, Luppino, and Rizzolatti (1985, 1991). (B) Penetration tracks reported on the outlines of three coronal sections passing through the recorded sectors. The dots plotted at different depths indicate sites at which tool-responding mirror neurons were found. c = central sulcus; ia = inferior arcuate sulcus; ip = intraparietal sulcus; l = lateral sulcus; p = principal sulcus; sa = superior arcuate sulcus; sp = spur of the arcuate sulcus; st = superior temporal sulcus.

with a higher capacity of generalizing the action goal. Another possible variable that could determine a Summing up, although F5 was originally thought of as an different neuron discharge during observation of actions homogeneous region, the cytoarchitectonic and func- made with the tool in respect to those performed with tional data open the possibility that it can be further the hand is represented by a difference in kinematics. In subdivided in different subregions. fact, the kinematics of actions made with the tool, The visual responses of tool-responding mirror neu- although reflecting that of the hand moving it, is differ- rons do not relate to different attention paid to tools in ent from that of the biological effector alone (unpub- respect to the biological effector or to a differential oc- lished data). On the other hand, the kinematics of ulomotor behavior. If the neuron response were related goal-directed and mimed actions made with the tool is to a higher level of attention given to the tool than to very similar. Thus, if kinematics were the crucial factor, the hand of the experimenter, we would expect to find the neuron should show a similar discharge in both a very high number of neurons activated by the sight of conditions. This is not what occurs. Tool-responding the action made with the tool and a smaller number of mirror neurons discharge only when the tool movement neurons responding to the observation of actions made is directed to a target. In addition, many tool-responding with the hand or the mouth. In contrast, we observed mirror neurons discharge when the tool, after having both types of neurons in the same recorded region. As reached the object, holds it, whereas they do not far as oculomotion is concerned, previous works dem- discharge when the same holding action is performed onstrated that the responses of F5 mirror neurons to with the hand. Because no movement is occurring in the action observation are typically not influenced by eye holding phase, kinematics cannot contribute to the behavior (Ferrari et al., 2003; Kohler et al., 2002). In neuron response. addition to this, it is known that neurons of area F5 are A further alternative explanation of the visual re- not related to the control of eye movements. sponse of tool-responding mirror neurons is that during

220 Journal of Cognitive Neuroscience Volume 17, Number 2 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/0898929053124910 by guest on 30 September 2021 observation, the monkey prepares to pick up the food tion of the biological effector (see also Maravita & Iriki, with the hand or with the mouth, planning to grasp the 2004). The same group, in a recent monkey PET study tool with different types of grip depending on the type (Obayashi et al., 2001), showed that in two monkeys of tool used. This explanation seems unlikely, because if trained to retrieve food with a rake, there was an this were true, the neuron response should have been activation of both parietal and ventral premotor cortex observed when food was simply presented to the mon- when the monkey used a rake to retrieve food, as key with the same tool used in the condition of action compared with a condition in which the monkey made observation. This was never observed. a simple manipulation of a stick. These findings suggest The visuomotor properties of tool-responding mirror that the motor training aimed to use a tool can change neurons resemble those of classical mirror neurons in not only the body schema (see Maravita & Iriki, 2004; Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/17/2/212/1757134/0898929053124910.pdf by guest on 18 May 2021 that they code the goal of an action. It has been Iriki et al., 1996) but also the motor representation of hypothesized that mirror neurons allow understanding body parts involved in goal-directed actions. Summing of action goal through a mechanism that matches action up, the link between the tool and the hand holding it observation with action execution (Gallese et al., 1996). endows the tool with the notion of the goal normally Note that this matching mechanism can reach a consid- associated to the biological effector. erable degree of abstraction. In fact, it has been recently Whether visual exposure to tool actions without demonstrated that one category of mirror neurons can motor training produce similar changes in the body recognize the goal of an action when the monkey only schema and/or in the motor representation of the hears the action sound (Kohler et al., 2002). Further- observer, as those shown by Iriki et al. (1996), has to more, it has been shown that several mirror neurons be determined. Comparing observational and motor become active during action observation also when the learning, the latter can rely on both the visual feedback final part of the observed action is occluded to the of the acting tool held by the hand and the kinesthetic monkey’s sight (Umilta` et al., 2001). Taken together, feedback, whereas the former relies only on a visual these studies indicate that the goal of the action can be feedback that is not directly correlated with a concurrent retrieved through different sensory inputs and suggest kinesthetic input. This difference is very important that the goal can be inferred also when only part of the because it could represent one of the reason why the visual information is provided. observed actions made with tools cannot be directly Keeping with this perspective, tool-responding mirror translated into own motor repertoire as also suggested neurons represent a new population of neurons with by our behavioral observations. the capacity to retrieve the action goal. For example, a A process of association between hand and tool neuron responding to the observation of a sticking and should be facilitated by a prolonged time of presentation holding action made with a stick respond also when the of the acting hand holding the tool. Our findings monkey grasps and holds a piece of food. Thus, the support this hypothesis in that tool-responding mirror visual (sticking) and motor (grasping) effective neuron neurons were discovered after several experimental responses signal the same general goal, that is taking sessions in both monkeys. In fact, in both monkeys, possession of an object. The fact that tool-responding these neurons have been found only in the last part of mirror neurons code a general action goal is also experiments, after months of training and recording reflected by their motor discharge. In fact, more than sessions. It is also interesting to note that in Monkey 70% of them respond equally well during the monkey’s 1, which was subjected to a longer experimental period execution of actions with the hand and the mouth, the than Monkey 2, there was a higher number of penetra- goal of the effective action being the same for both tions in which tool-responding mirror neurons were effectors. found compared with that of Monkey 2. Furthermore, How can a movement toward a target performed with the increase in frequency of tool-responding mirror a tool be recognized as goal-directed? A possible expla- neurons in the last penetrations of Monkey 1 suggests nation is that after a relatively long exposure to actions that the period of exposure to actions made with tools made with tools, a visual association between the tool can be an important factor to facilitate the association and the hand is created, so that it becomes as a kind of between hand and tool. Despite these observations, prolongation of, or even a surrogate for, the hand. The further studies are needed to directly specify the rela- association between the two agents would facilitate the tionship between different times of exposure to tool generalization of the action meaning. The possibility of actions and the emergence of tool-responding mirror the existence of cortical mechanisms able to represent a neurons. tool as a prolongation of the arm was already proposed The process of generalization of the action goal by Iriki, Tanaka, and Iwamura (1996), who demonstrated between the hand and the tool can be helped by several in monkeys that after a motor training to use a tool, the factors: (1) The experimenter in both cases is making properties of forelimb-related neurons in the parietal movements that end on a target; (2) the general kine- cortex could be extended to tools held by the forelimb, matic properties are similar. For example, in both ac- as if these tools would constitute the natural prolonga- tions made with the tool and with the hand, there is an

Ferrari, Rozzi, and Fogassi 221 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/0898929053124910 by guest on 30 September 2021 acceleration and a deceleration phase, followed by a Kummer, 1996; Tokida et al., 1994; Beck, 1976). These static phase in which the effector is in contact with the findings could explain why the monkey used in our target; (3) there is a similarity of the context (same experiment did not use the stick to retrieve food. The environment, same experimenters) in which the actions monkey probably would have needed a longer time of take place; (4) experimenter’s hand and tools can both learning in which the observation of another individual be associated, by the monkey, to the possibility to using tool were combined with its own sensorimotor receive food. In fact, both hand and tools were used interaction with the same tool. during the experiments to give food to the monkeys, In the light of the present findings and of the above sometimes, also during action-observation conditions. In considerations, what could be the function of tool- particular, the stick was the tool mostly used for feeding responding mirror neurons? They probably are part of Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/17/2/212/1757134/0898929053124910.pdf by guest on 18 May 2021 the monkey. In agreement with this observation, there a system enabling the observing individual to extend its were more tool-responding mirror neurons responding action understanding capacity to actions that are not best to the observation of actions made with the stick strictly part of its motor representation. Hence, this than with the pliers. extension would include also acting objects. Accepting the idea that the tool can acquire the same It is common observation also in humans that it is meaning of the acting hand, the issue is raised on why possible to understand the meaning of an action without neurons of premotor cortex can respond, visually, better possessing the immediate capacity to reproduce it. In to actions made with tools than to actions made with the agreement with this, a recent fMRI experiment in hu- hand. The hypothesis we propose is that different visual mans (Buccino, Vogt, et al., 2004) demonstrates that inputs related to stimuli with similar biological motion the pure observation of a sequence of actions (playing and directed to the same target object can access F5 guitar chords) that were not familiar to the observing motor neurons that on the motor side encode mouth subjects determines an activation of the inferior frontal and hand actions having the same goal, but on the visual gyrus (Broca’s area), a region considered homologue to side are probably still uncommitted. The reason why monkey’s area F5. This supports the notion that pre- tool-responding mirror neurons prefer tools and mirror motor areas in primates can be involved in processes neurons prefer biological effectors can be attributed to a of abstraction of the goal aimed to understand actions competition mechanism between two visual inputs for also when the individual has never experienced them or the same population of uncommitted neurons, which in when he has not yet acquired the capacity to master some cases, after long visual exposure, become domi- them. nated by one input that is not a biological effector. This possibility seems not unlikely because the monkeys used Conclusions in this experiment had not only the opportunity to see actions made with biological effectors (both by conspe- The large use of tools in primates and the diversity of cifics and human experimenters) but also to observe their use among the different populations suggest the very frequently actions made with tools. An extreme case existence of neural mechanisms not only aimed to of the proposed competition mechanism can be man- understand actions made with tools but also to learn ifested by the neuron shown in Figure 5, in which their use and to transmit the new acquired skills to other observation of a sticking action elicits a neuron excita- individuals. Studies made in apes have suggested that tion, whereas observation of a hand-grasping action this transmission can be achieved through observational produces an inhibitory response. learning processes such as imitation (van Schaik et al., Ethological studies show that monkeys are able to 2003; Whiten et al., 1999). In contrast, imitation in use tools in captivity or in the wild (Tokida, Tanaka, monkeys has not been demonstrated to occur, and this Takefushi, & Hagiwara, 1994; Westergaard, 1988; Beck, can account for the lack of population-wide tool use in 1976; see Tomasello & Call, 1997, for a review). In ma- the wild. However, it is very likely that the neurophys- caques, tool use is a behavior that can be often observed iological mechanisms that lead to the wide use of tools by group members both in the wild and in captivity in apes and humans rely on the action understanding (Tomasello & Call, 1997), although it is usually not very system, described above and in previous studies (see widespread within each wild population concerned (i.e., Rizzolatti, Fogassi, & Gallese, 2001), already present in shown by many different individuals, regularly or pre- monkey’s premotor area F5. The involvement of this dictably) (van Schaik, Deaner, & Merrill, 1999). The fact area during monkey’s tool use (Obayashi et al., 2001) that tool use in monkeys rarely becomes habitual or supports this hypothesis. In an evolutionary perspective, customary in a given population is in agreement with the it is not the first time that the emerging properties of a finding that a relatively long period is necessary before a brain area can favor the evolution of new functions. This new skill with the tool is acquired (see Tomasello & Call, view is congruent with the increasing evidence that a 1997). As some authors have proposed, the acquisition mirror mechanism, which evolved in monkeys for action of tool use requires processes involving observational understanding, emerged subsequently in human evolu- and individual learning (Zuberbu¨hler, Gygax, Harley, & tion as suitable neural substrate for imitation.

222 Journal of Cognitive Neuroscience Volume 17, Number 2 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/0898929053124910 by guest on 30 September 2021 METHODS cortex). For the criteria used to functionally characterize General Procedures these areas, see the work of Umilta` et al. (2001). Once the location of these areas was characterized, we con- The experiments were carried out on two awake (Mon- centrated our recordings on area F5. key 1 and Monkey 2), partially restrained macaque monkeys (M. nemestrina). All experimental protocols were approved by the Veterinarian Animal Care and Use Neuron Testing Committee of the University of Parma as well as by the Each neuron, once isolated, was tested to ascertain its Italian Minister of Health. All experiments complied with motor and visual properties. The type of testing is the European law on the humane care and use of described below. Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/17/2/212/1757134/0898929053124910.pdf by guest on 18 May 2021 laboratory animals. Before starting the single-neuron recording experi- ments, the monkeys were habituated to the experiment- Motor Properties ers and to the experimental conditions. Each monkey Neurons discharging in association with active hand was seated on a primate chair and habituated to observe actions were tested using the same stimuli and proce- different types of visual stimuli and to execute different dures as in previous studies. Briefly, the monkey was types of hand and mouth actions. During this training presented with a variety of objects of different size and period, both monkeys could see, among other visual shape. They consisted of food items and objects at hand stimuli, the tools used in the present experiment. The in the laboratory. The objects were presented in differ- stick was used both for picking up food and for present- ent parts of space within the monkey reaching distance. ing and giving pieces of food to the monkey. Pliers were The monkey was trained to reach for and grasp them mainly used for grasping food or objects and giving with different types of grip (for details, see Rizzolatti them to the monkey. In general, the monkeys were et al., 1988, 1990). Neurons discharging in association more exposed to stick than to pliers. The monkey with mouth actions were studied by giving the monkey training lasted approximately 2 months. food or juice, thus eliciting a variety of mouth actions The surgical procedures for single-neuron recordings (see Ferrari et al., 2003). were the same as previously described (Fogassi et al., 1996). The head implant included a head holder and a chamber for single-unit recordings. Neurons were re- Visual Properties corded using tungsten microelectrodes (impedance 0.5– Testing of visual properties consisted in the presentation 1.5 M , measured at 1 kHz) inserted through the dura of actions performed with the hand, the mouth, and a that was left intact. Neuronal activity was amplified and tool by the experimenter in front of the monkey. The monitored on an oscilloscope. Individual action poten- tested hand actions were object grasping, holding, ma- tials were isolated with a dual voltage–time window nipulating, breaking, and tearing. The tested mouth discriminator (Bak Electronics, Germantown, MD). The actions were ‘‘grasping,’’ ‘‘holding,’’ ‘‘breaking,’’ ‘‘suck- output signal from the voltage–time discriminator was ing,’’ ‘‘reaching with the tongue,’’ ‘‘taking away,’’ and monitored and fed to a PC for analysis. ‘‘chewing.’’ Movements of the hand or the mouth mim- The recording microelectrodes were used also for ing actions in the absence of the target were also electrical microstimulation (train duration, 50 msec; presented (for other details, see Ferrari et al., 2003; pulse duration, 0.2 msec; frequency, 330 Hz; current Gallese et al., 1996). intensity, 3–40 AA). The current strength was controlled The tool actions were performed by the experimenter on an oscilloscope by measuring the voltage drop across with a wooden stick (length 37 cm) having a metal tip on a 10-K resistor in series with the stimulating electrode. one extremity and a pair of pliers. The actions per- formed with the stick were sticking, holding, manipulat- ing, and taking away. (A) Sticking. The food, held on the Recording Sites experimenter’s hand or on a tray, was approached by The chamber for single-unit recordings was implanted the stick held by the experimenter with his other hand stereotactically. The chamber rostro-caudal and medio- and then punctured with the tip of the stick. (B) lateral axes dimension (20 Â 15 mm) was such as to allow Holding. Once the food was punctured, it was held on to record from the whole ventral premotor cortex. The the stick without subsequent movement. (C) Manipu- chamber covered a region from area F1 (primary motor lating. After reaching the food, the stick was moved cortex) posteriorly to the caudal part of the frontal eye around a piece of food. During this action the food was fields anteriorly. After chamber implantation, the ventral touched but not punctured. (D) Taking away. Once the part of the agranular frontal cortex was functionally food was punctured, it was held on the stick tip and explored (single-neuron recordings and intracortical then taken away from the tray or from the supporting microstimulation) to assess the location of areas F1 hand. The actions performed with the pliers were (primary motor cortex), F4, and F5 (ventral premotor grasping, holding, manipulating, breaking, and taking

Ferrari, Rozzi, and Fogassi 223 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/0898929053124910 by guest on 30 September 2021 away. (A) Grasping. The experimenter, holding the Epoch 1 corresponds to the time interval (1 sec) pliers with one hand, approached a piece of food held centered at the alignment signal (500 msec before and by a support or by the other hand and grasped it by 500 msec after it) and Epoch 2 corresponds to a period opposing the tips of the pliers. (B) Holding. Once the of 2000 msec to 1000 msec before alignment (back- food was grasped, it was held with the pliers without ground activity). To compare, for each neuron, the subsequent movement. (C) Manipulating. After reach- discharge in different epochs and conditions two-way ing the food, the pliers were moved around a piece of ANOVA (2 Â 2 ANOVA, factors condition and epoch, food, touching it but without grasping it. (D) Breaking. two levels each) was performed. ANOVA was followed The experimenter, holding the pliers with one hand, by Newman–Keuls post hoc tests comparisons to com- approached a piece of food held by a support or by his pare the neuron activity recorded during observation of Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/17/2/212/1757134/0898929053124910.pdf by guest on 18 May 2021 other hand and broke it employing force with the tips of actions performed with the tool with the activity re- the pliers. (E) Taking away. Once the food was corded during observation of actions made with the grasped, it was held between the tips of the pliers and hand, the mouth, or during simple presentation. All then taken away from the support/hand. analyses were performed using a significance criterion To assess the specificity of the neuron response to the of p < .05. observation of actions performed with the tool, move- ments of the stick or of the pliers miming actions without the target object were presented. In some Histological Control neurons, the specificity of the response was further Histology was performed in both monkeys. For detailed tested using elongated tools similar to the stick (i.e., a description of the histological processing, see Fogassi, signal stick, which is a 15-cm diameter, wooden, colored Gallese, Buccino, Craighero, and Rizzolatti (2001). Brief- circle attached to the tip of a 1-m long stick) to ly, each animal was anaesthetized with ketamine hydro- approach, to touch, and to go away from the target. chloride (15 mg/kg im) followed by an intravenous Note that these tools could not be used to grasp or take lethal injection of pentobarbital sodium and perfused possession of objects. through the left cardiac ventricle with buffered saline, All of the above-described actions were performed by followed by fixative (paraformaldehyde 3.5%). After the an experimenter standing in front of the monkey. The monkey was sacrificed, the brain was removed from the actions were presented to the monkeys mostly by two skull, photographed, and then frozen and cut coronally. experimenters. Although the two monkeys were partic- The sections (60 Am thick) were stained using the Nissl ularly familiar with the experimenter who trained the method. For each section, outer and inner cortical monkey before recording, no obvious difference in contours were drawn. The locations in the cortex of neuron response related to which experimenter per- the electrode tracks were assessed under optic micro- formed the action was observed. scope and then plotted and digitalized using an Inside Other visual stimuli such as static or moving objects PC program. The tracks location was subsequently relat- were presented on a tray or on a stick held by an ed to the areas of the ventral premotor cortex. For each experimenter, in all parts of the visual field, inside and electrode track, the depth at which tool-responding outside the monkey’s reaching distance. mirror neurons were found was reported. A second PC program (CRS4, Cagliari, Italy, see Bettio, Demelio, Data Analysis Gobbetti, Luppino, & Matelli, 2001) was then used for 3-D reconstruction of the brain, which allowed also brain After neuron visual and motor characterization, histo- reslicing in a different plane of cutting. grams of visual and motor responses were constructed. By using a contact-detecting circuit (a switch), a signal was sent to a PC whenever the monkey (motor con- Acknowledgments ditions) or the experimenter (visual conditions) touched We thank G. Rizzolatti for his valuable comments on an early a metal surface, connected to the target, with their hand, draft of the manuscript. We thank G. Luppino for his help in their mouth, or with the tool. This signal was used to discussing anatomical data, A. C. Roy for her help in discussing kinematics aspects, and R. Romani for his valuable technical align the trials and, subsequently, to construct the support. This work was supported by MIUR and Italian CNR. response histograms. Switch closing produced a signal that was used for the response alignment. Response Reprint request should be sent to: Leonardo Fogassi, PhD, Dipartimento di Neuroscienze, Sezione Fisiologia, Universita` di histograms were constructed by summing 10 individual Parma, via Volturno 39, 43100 Parma, Italy, or via e-mail: trials. Unless some technical problems did not interfere [email protected]. with trial acquisition (e.g., delays in closing the contact- detecting circuit), no trials were discarded. To statistically assess the neuron discharge, the neu- REFERENCES ronal activity, expressed as mean firing rate (spikes per Arbib, M. A., & Rizzolatti, G. (1999). Neural expectations. second), was measured in two different time epochs. A possible evolutionary path from manual skills to language.

224 Journal of Cognitive Neuroscience Volume 17, Number 2 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/0898929053124910 by guest on 30 September 2021 In P. Van Loocke (Ed.), The nature of concepts. Maravita, A., & Iriki, A. (2004). Tools for the body (schema). Evolution, structure and representation (pp. 128–154). Trends in Cognitive Sciences, 8, 79–86. New York: Routledge. Matelli, M., Luppino, G., & Rizzolatti, G. (1985). Patterns of Barbas, H., & Pandya, D. N. (1987). Architecture and frontal cytochrome oxidase activity in the frontal agranular cortex of cortical connections of the premotor cortex (area 6) in the the macaque monkey. Behavioral Brain Research, 18, rhesus monkey. Journal of Comparative Neurology, 256, 125–136. 211–228. Matelli, M., Luppino, G., & Rizzolatti, G. (1991). Architecture of Beck, B. B. (1976). Tool use by captive pigtail macaques. superior and mesial area 6 and the adjacent cingulate Primates, 17, 301–310. cortex in the macaque monkey. Journal of Comparative Bettio, F., Demelio, S., Gobbetti, E., Luppino, G., & Matelli, M. Neurology, 311, 445–462. (2001). Interactive 3-D reconstruction and visualization of Nishitani, H., & Hari, R. (2000). Temporal dynamics of cortical primates cerebral cortex. Society for Neuroscience Abstracts, representation of action. Proceedings of the National Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/17/2/212/1757134/0898929053124910.pdf by guest on 18 May 2021 728.4. Academy, U.S.A., 97, 913–918. Brodmann, K. (1905). Beitrage zur histologischen lokalisation Obayashi, S., Suhara, T., Kawabe, K., Okauchi, Maeda, J., der grosshirnrinde. III. Mitteilung: Die rindenfelder der Akine, Y., Onoe, H., & Iriki, A. (2001). Functional niedere affen. Journal fu¨r Psychologie und Neurologie, 4, brain mapping of monkey tool use. Neuroimage, 14, 177–226. 853–861. Buccino, G., Binkofski, F., Fink, G. R., Fadiga, L., Fogassi, L., Petrides, M., & Pandya, D. N. (1994). Comparative architectonic Gallese, V., Seitz, R. J., Zilles, K., Rizzolatti, G., & Freund, H. J. analysis of the human and the macaque frontal cortex. (2001). Action observation activates premotor and parietal In F. Boller & J. Grafman (Eds.), Handbook of areas in a somatotopic manner: An fMRI study. European neuropsychology (pp. 17–58). Amsterdam: Elsevier. Journal of Neuroscience, 13, 400–404. Preuss, T. M., & Goldman-Rakic, P. S. (1991). Myelo- and Buccino, G., Vogt, S., Ritzl, A., Fink, G. R., Zilles, K., Freund, cytoarchitecture of the granular frontal cortex and H. J., & Rizzolatti, G. (2004). Neural circuits underlying surrounding regions in the strepsirhine primate Galago and learning of hand actions: An event-related fMRI study. the anthropoid primate Macaca. Journal of Comparative Neuron, 42, 323–334. Neurology, 310, 429–474. di Pellegrino, G., Fadiga, L., Fogassi, L., Gallese, V., & Rizzolatti, G., Camarda, R., Fogassi, L., Gentilucci, M., Rizzolatti, G. (1992). Understanding motor events: A Luppino, G., & Matelli, M. (1988). Functional organization of neurophysiological study. Experimental Brain Research, 91, inferior area 6 in the macaque monkey: II. Area F5 and 176–180. the control of distal movements. Experimental Brain Ferrari, P. F., Fogassi, L., Gallese, V., & Rizzolatti, G. (2001). Research, 71, 491–507. Mirror neurons for mouth actions in the monkey ventral Rizzolatti, G., Fadiga, L., Gallese, V., & Fogassi, L. (1996). premotor cortex. Society for Neuroscience Abstracts, 729.4. Premotor cortex and the recognition of motor actions. Ferrari, P. F., Gallese, V., Rizzolatti, G., & Fogassi, L. (2003). Cognitive Brain Research, 3, 131–141. Mirror neurons responding to the observation of ingestive Rizzolatti, G., Fadiga, L., Matelli, M., Bettinardi, V., Paulesu, E., and communicative mouth actions in the monkey ventral Perani, D., & Fazio, F. (1996). Localization of grasp premotor cortex. European Journal of Neuroscience, 17, representations in humans by PET: 1. Observation versus 1703–1714. execution. Experimental Brain Research, 111, 246–252. Fogassi, L., Gallese, V., Buccino, G., Craighero, L., & Rizzolatti, Rizzolatti, G., Fogassi, L., & Gallese, V. (2000). Cortical G. (2001). Cortical mechanism for the visual guidance of mechanisms subserving object grasping and action hand grasping movements in the monkey: A reversible recognition: A new view on the cortical motor function. inactivation study. Brain, 124, 571–586. In M. S. Gazzaniga (Ed.), The new cognitive neurosciences Fogassi, L., Gallese, V., Fadiga, L., Luppino, G., Matelli, M., & (2nd ed., pp. 539–552). Cambridge: MIT Press. Rizzolatti, G. (1996). Coding of peripersonal space in inferior Rizzolatti, G., Fogassi, L., & Gallese, V. (2001). premotor cortex (area F4). Journal of Neurophysiology, 76, Neurophysiological mechanisms underlying the 141–157. understanding and imitation of action. Nature Neuroscience Gallese, V., Fadiga, L., Fogassi, L., & Rizzolatti, G. (1996). Reviews, 2, 661–670. Action recognition in the premotor cortex. Brain, 119, Rizzolatti, G., Gentilucci, M., Camarda, R., Gallese, V., Luppino, 593–609. G., Matelli, M., & Fogassi L. (1990). Neurones related to Grafton, S. T., Arbib, M. A., Fadiga, L., & Rizzolatti, G. (1996). reaching-grasping arm movements in the rostral part of Area Localization of grasp representations in humans by positron 6 (Area 6 a b). Experimental Brain Research, 82, emission tomography. 2. Observation compared with 337–350. imagination. Experimental Brain Research, 112, Roberts, T. S., & Akert, K. (1963). Insular and opercular cortex 103–111. and its thalamic projections in Macaca mulatta. Archiv Gre`zes, J., Costes, N., & Decety, J. (1998). Top down effect fu¨r Neurologie, Neurochirurgie und Psichiatrie, 92, 1–43. of the strategy on the perception of human biological Tokida, E., Tanaka, I., Takefushi, H., & Hagiwara, T. motion: A PET investigation. Cognitive Neuropsychology, 15, (1994). Tool-using in Japanese macaques: Use of stones to 553–582. obtain fruit from a pipe. Animal Behaviour, 47, Iacoboni, M., Woods, R. P., Brass, M., Bekkering, H., Mazziotta, 1023–1030. J. C., & Rizzolatti, G. (1999). Cortical mechanisms of Tomasello, M., & Call, J. (1997). Primate cognition. Oxford: human imitation. Science, 286, 2526–2528. Oxford University Press. Iriki, A., Tanaka, M., & Iwamura, Y. (1996). Coding of modified Umilta`, M. A., Kohler, E., Gallese, V., Fogassi, L., Fadiga, L., body schema during tool use by macaque postcentral Keysers, C., & Rizzolatti, G. (2001). I know what you are neurones. NeuroReport, 7, 2325–2330. doing a neurophysiological study. Neuron, 31, 155–165. Kohler, E., Keysers, C., Umilta`, M. A., Fogassi, L., Gallese, V., van Schaik, C. P., Ancrenaz, M., Borgen, G., Galdikas, B., Knott, & Rizzolatti, G. (2002). Hearing sounds, understanding C. D., Singleton, I., Suzuki, A., Utami, S. S., & Merrill, M. actions: Action representation in mirror neurons. Science, (2003). Orangutan cultures and the evolution of material 297, 846–848. culture. Science, 299, 102–105.

Ferrari, Rozzi, and Fogassi 225 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/0898929053124910 by guest on 30 September 2021 van Schaik, C. P., Deaner, R. O., & Merrill, M. Y. (1999). The silenus) manufacture and use tools. Journal of conditions for tool-use in primates: Implications for the Comparative Psychology, 102, 152–159. evolution of material culture. Journal of Human Evolution, Whiten, A., Goodall, J., McGrew, W. C., Nishida, T., 36, 719–741. Reynolds, V., Sugiyama, Y., Tutin, C. E., Wrangham, R. W., Vogt, C., & Vogt, O. (1919). Allgemeinere ergebnisse & Boesch, C. (1999). Cultures in chimpanzee. Nature, 399, unserer hirnforschung. Journal fu¨r Psychologie und 682–685. Neurologie, 25, 279–462. Zuberbu¨hler, K., Gygax, L., Harley, N., & Kummer, H. (1996). von Bonin, G., & Bailey, P. (1947). The neocortex of Macaca Stimulus enhancement and spread of a spontaneous tool mulatta. Urbana: University of Illinois Press. use in a colony of long-tailed macaques. Primates, 37, Westergaard, G. C. (1988). Lion-tailed macaques (Macaca 1–12. Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/17/2/212/1757134/0898929053124910.pdf by guest on 18 May 2021

226 Journal of Cognitive Neuroscience Volume 17, Number 2 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/0898929053124910 by guest on 30 September 2021